loading...
Lancey runs AI micro-experiments on precise behavioral segments, not generic cohorts – turning personalization into infrastructure rather than a feature toggle.
ENTRY ANGLES
AI-powered user behavior analysis to tailor experiences per segment · Infrastructure platforms enabling other product teams to implement personalization · Personalization focused on activation, engagement, and retention metrics
VERTICALS
CAPABILITIES
AI/machine learning for user behavior analysis, Platform infrastructure and APIs, Personalization engine development
LANCEY FOUNDER
“everyone on the Pro plan.”
Most A/B tests treat users as a uniform mass. Lancey bets that the interesting work happens at a much finer grain: AI-automated micro-experiments that target precisely defined behavioral segments, not generic "all Pro users" buckets.
A micro-experiment targets a small, precisely defined group of users – selected not just by demographic attributes but by past behavior patterns. That's what sets it apart from traditional product experiments, which typically run across all users or large segments like "everyone on the Pro plan."
Why bother with micro-experiments? Because users are not a monolith. They behave differently, they use features differently, and what drives engagement for one cohort may actively alienate another. There's no single universal lever that simultaneously improves activation, engagement, and retention for everyone.
Manually analyzing every user's behavior, splitting them into meaningful sub-groups, designing experiments per group, and measuring results is a monumental hassle. Lancey is built to remove that bottleneck.
Setup begins by connecting all available behavioral data sources to Lancey – not just in-product signals, but everything around the product too. Data can flow in from Segment, Snowflake, Mailchimp, payment processors, and similar tools. Pre-built integrations cover the most popular sources.
Next, a product team sets a goal – improve daily active use, boost feature engagement, reduce churn, or something else. Lancey's AI then automatically identifies which user attributes and behavioral patterns most strongly influence that metric.
From there, the AI segments all users based on those patterns and characteristics. Then it generates a ranked list of experiments – each targeting a small sample from each segment – so teams can measure impact against a control group from the same segment.
Each experiment requires a coded procedure – say, displaying a banner with a specific call to action. The platform then calls that procedure for whichever sample group needs it.
Over time, running individual experiments builds a library of "nudges." Once that library has enough coverage, Lancey can shift into autopilot mode: the AI continuously analyzes user behavior, segments users, designs and runs experiments using the existing library, measures outcomes, and rolls out the winning nudge combinations to entire segments – all without human involvement.
Lancey went through Y Combinator in summer 2022, receiving the standard $500K. The beta platform launched only a few days before this writing.
Most businesses still operate on the "one size fits all" principle – not out of malice, but because the time, money, and energy required to handle every customer individually simply isn't there.
Digital products face the same reality. Digging into individual user behavior, understanding what drives it, and connecting that behavior to product outcomes is painstaking work that most product teams can't afford. It's faster and easier to swing a broad axe – making product and marketing decisions aimed at the "average user" and hoping the results follow.
But that approach leaves money on the table. The "long tail" of users whose behavior doesn't match the assumed average gets systematically ignored. The wider the product's feature set, the longer that tail grows – because different users use different features toward different ends, and tailoring the experience to all of them becomes exponentially harder.
Even simple products see users fragment into distinct behavioral patterns. A [related review](/review/privychka-vazhnee-chem-polza) previously explored how even readers of a niche newsletter access it in starkly different ways – some on the web, some via Telegram, some every morning one post at a time, some in seven-post batches on Wednesday evenings or Sunday afternoons.
As a result, trying to push all subscribers toward a single engagement pattern would actively hurt most of them. That review [covered](/review/privychka-vazhnee-chem-polza) Subsets – a platform that helps subscription publishers reduce churn by segmenting readers by behavioral pattern and identifying the right retention action for each group.
Lancey is conceptually similar to Subsets, but generalizes beyond digital media to any product type.
OfferFit, [covered previously](/review/kak-povysit-jeffektivnost-reklamnyh-rassylok) in November, operates along the same lines. They built a platform that sends personalized purchase offers to customers based on individual preferences – what they buy most, what messaging resonates, even what time of day they're most likely to open an email or text. The result: every customer receives a unique offer, framed uniquely, delivered at the individually optimal moment.
Lancey, Subsets, and OfferFit all represent what could be called "scalable personalization" – the ability to adapt to each user individually, at scale, without the cost exploding. Segmenting into small behavioral groups is essentially a practical approximation of true one-to-one personalization: people are unique, but some are more alike than others.
This has only become possible because of AI – which can now perform meticulous user analysis and continuous adaptation in fully automated fashion.
The concept works well beyond subscriptions or e-commerce. An [earlier review](/review/individualno-i-so-100-rezultatom) covered Studio – a well-funded online course platform ($60M raised) that recently launched what it calls "the music school of the future," promising every enrolled student will produce one polished, publishable song per month. The secret weapon: an AI instructor that assembles a personalized curriculum for each student and provides individualized coaching throughout. The instructor also groups students into cohorts of roughly 20 by skill level and goal similarity. Without that AI layer, the promise would be economically impossible to keep.
The direction is clear: toward scalable personalization. The problem product teams face is well understood, and AI-powered solutions to it have only recently become viable.
On the product side: use AI to deeply analyze your own users' behavior and tailor the experience per segment – improving activation, engagement, and retention. On the infrastructure side: build platforms that let other product teams do the same.
Even the handful of startups mentioned here show how many different angles exist. Subscription services, e-commerce, and education are the obvious starting points – and still large enough to support multiple successful entrants.
But the more interesting question is: where else do users, customers, or buyers naturally fragment into distinct behavioral patterns while using the same product? And where can adapting to those patterns meaningfully lift revenue for the product creator?